Proposed SD Algorithm Defective

[Blake Cretney]

--

On Wed, 19 Aug 1998 00:15:48   Norman Petry wrote:
>With a couple of slight changes, the algorithm idea I posted a few days ago
>is _close_ to being an accurate implementation of SD.  Here's how the steps
>should be listed:
>
>1) Check for any unbeaten candidates.  If 1 or more, then quit.
>2) Repeatedly eliminate beat-nothing candidates (E, then D in example).
>3) Break the weakest defeat(s).  Goto  step 1.
>

I think the best bet for an SD algorithm is
Repeat
    Eliminate everything not in the Simpson set
    Break lowest defeat
Until only one candidate is left

If anyone wants pseudo-code for doing this (including the Simpson set elimination step), I could provide it.

I wouldn't describe this as simple, but it is possible, and it is more important that a method be simple to describe than to impliment.


-----== Sent via Deja News, The Discussion Network ==-----
http://www.dejanews.com/  Easy access to 50,000+ discussion forums